Token-Based Collaborative Machine Learning for Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Entities often face challenges in sharing information for collaborative machine learning due to security, regulatory, or contractual constraints, leading to sub-optimal decision-making and impaired performance in applications like cybersecurity risk assessment.
Innovation Solution
A token-based approach where a first machine learning model outputs tokens that do not reveal more than a threshold amount of information, allowing these tokens and associated token-context values to be shared with a second entity, enabling collaborative machine learning while preserving confidentiality and compliance with privacy regulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If entities share information for collaborative machine learning, then machine learning accuracy is improved, but confidentiality and compliance are compromised
Solution Approach 1:
The patent introduces an intermediary tokenization mechanism that transforms sensitive data into tokens before sharing between entities. The tokenization layer acts as a mediator that preserves the utility of data for machine learning while preventing direct access to confidential information, thus resolving the contradiction between accuracy improvement and confidentiality protection
Solution Approach 2:
The patent creates tokenized copies of sensitive data that can be shared for machine learning purposes without exposing the original confidential information. These token copies maintain the statistical and relational properties needed for accurate machine learning while being inert and non-sensitive, thus enabling accuracy improvement without confidentiality breach
2Productivity
If entities share information for collaborative machine learning, then decision-making quality is improved, but regulatory compliance is violated
Solution Approach 1:
The tokenization mechanism serves as an intermediary that enables information sharing for improved decision-making while maintaining regulatory compliance. By transforming data into tokens before transmission, the system allows collaborative machine learning to produce higher quality decisions without violating regulatory requirements regarding data privacy and security
Solution Approach 2:
The patent changes the parameter state of data from sensitive to non-sensitive through tokenization. This parameter transformation allows the data to be shared for collaborative decision-making while maintaining regulatory compliance, as the tokenized form no longer contains personally identifiable or sensitive information
3Reliability
If entities restrict information sharing due to security constraints, then confidentiality is maintained, but machine learning performance is impaired
Solution Approach 1:
The patent creates tokenized copies of confidential data that can be shared with machine learning models without compromising the original confidential information. These token copies preserve the necessary data characteristics for accurate machine learning performance while the original data remains protected, thus maintaining confidentiality without impairing performance
4Reliability
If entities restrict information sharing due to contractual constraints, then confidentiality is maintained, but collaborative decision-making is sub-optimal
Solution Approach 1:
The tokenization mechanism acts as an intermediary that enables collaborative decision-making while respecting contractual constraints on information sharing. By transforming data into tokens, the system allows entities to collaborate and produce optimal decisions without directly sharing confidential information, thus maintaining confidentiality while improving collaborative productivity
Data Source
AI summary
Provided is a process that includes sharing information among two or more parties or systems for modeling and decision-making purposes, while limiting the exposure of details either too sensitive to share, or whose sharing is controlled by laws, regulations, or business needs.